Let us assume that the series follows the general model that can be rewritten in terms of the present and past values of :
Our objective is to forecast a future value given our information set that consists of the past values . The future value is generated by model (4.32), thus
Let us denote by the -step ahead forecast of
made at origin . It can be shown that, under reasonable weak conditions,
the optimal forecast of
is the conditional expectation of
given the information set, denoted by
.
The term optimal is used in the sense that minimizes the Mean Squared Error
(MSE). Although the conditional expectation does not have to be a linear
function of the present and past values of , we shall consider linear
forecasts because they are fairly easy to work with. Furthermore, if the
process is normal, the Minimum MSE forecast (MMSE) is linear. Therefore,
the optimal forecast -step ahead is:
The -step ahead forecast error is a linear combination of the future shocks entering the system after time :
Given these results, if the process is normal, the forecast interval is:
For , the one-step ahead forecast error is , therefore can be interpreted as the one-step ahead prediction error variance.
Let us consider again the general model that can be written as well as:
In practice, the parameters of the model should be estimated, but for convenience, we assume that they are given.
Following the results of section 4.4.2, if the series follows an model, the -step ahead forecast at origin is given by:
The expression (4.37) is called the eventual forecast function, because it holds only for . If , then the eventual forecast function holds for all . This eventual forecast function passes through the values given by .
Let us consider the process in deviations to the mean :
Let us take as an example the forecast of an process with and . Figure 4.12 shows the eventual forecast function of the model considered (dotted line). It can be observed that this function increases at the beginning until it reaches the mean value, 3. This result, that is,
Let us consider the following model:
If , we get the random walk model (4.20) and the eventual forecast function takes the form:
The eventual forecast function for the random walk plus drift model (4.21) is the solution to the following difference equation:
It can be observed in figure 4.13 that the interval forecast limits increase continuously as the forecast horizon becomes larger. It should be taken into account that when the process is nonstationary the limit does not exist.
Let us consider the model: